Anthropic is planning to issue super-voting shares to its CEO and co-founders before its IPO. This is not just a governance detail—it’s a signal about the future of AI capitalism. The announcement, first reported by Crypto Briefing, comes as the AI lab races toward a public listing that could value it at hundreds of billions. But the structure they’re choosing carries a heavy cost: a built-in governance discount that could alienate institutional investors and inflame regulatory scrutiny.
Let’s start with the facts. Anthropic, the company behind the Claude model series, has raised over $10 billion from strategic investors like Amazon and Google. Its mission-driven ethos—focusing on AI safety and long-term alignment—has been a key differentiator in the crowded LLM market. Now, ahead of its IPO, the founders want to lock in control. The tool: super-voting shares, typically granting 10x or 20x voting power per share compared to the common stock offered to the public. This is a classic dual-class structure, used by Google, Meta, and Snap. But in the context of AI, it’s a minefield.
The numbers don’t lie. Research from Institutional Shareholder Services (ISS) and Glass Lewis shows that dual-class companies trade at a discount of 5% to 10% relative to single-class peers. The discount is larger for firms with weaker sunset provisions or younger founders. For a company like Anthropic, which is already valued at $60 billion+ in private markets, a 10% discount translates to billions in lost market cap. That’s the price of control.
But here’s the core insight: this discount is not uniform. It depends on the reason for the control structure. In the case of Alphabet, the dual-class structure was justified by the need for long-term vision in advertising and search. For Meta, it was about Zuckerberg’s unassailable control. For Anthropic, the narrative is different. They claim super-voting shares are a “mission protection mechanism”—a way to ensure that short-term profit pressures don’t dilute the company’s commitment to AI safety. In my years analyzing cross-border payment systems, I’ve seen similar control structures throttle innovation when they become too rigid. But in AI, the stakes are existential. The question is whether the market will buy that narrative.
The code doesn’t care about your narrative. The market is a cold, calculating machine. It looks at the balance sheet, the revenue multiples, the governance structure. If Anthropic’s super-voting shares have no sunset clause—meaning the control lasts indefinitely or until the founders step down—investors will demand a higher risk premium. That premium will manifest in a lower IPO price, a smaller public float, and a narrower investor base. Index funds and ESG-focused funds, which are increasingly sensitive to governance, may simply skip the stock. That’s not a theoretical risk. In 2023, BlackRock and Vanguard voted against the re-election of directors at several dual-class companies, citing lack of accountability.
But here’s where the contrarian angle comes in. What if the governance discount is actually smaller for AI companies because of the unique nature of the technology? The reasoning is straightforward: AI is a winner-take-most market where the quality of the model and the speed of iteration matter more than quarterly earnings. A distracted board or activist investors pushing for layoffs could kill the very product that makes the company valuable. In that context, concentrated control might be a competitive advantage. Snap’s dual-class structure didn’t stop it from growing—it just made its stock more volatile. And Google’s founders used their control to make billion-dollar bets on machine learning that paid off handsomely.
The market’s current capabilities are already pricing in this narrative. Anthropic’s private valuation has held steady despite the governance rumors. That suggests that strategic investors like Amazon and Google—who see the company as a key supplier of AI models for their cloud businesses—are comfortable with the arrangement. They’re not buying voting rights; they’re buying access to the technology. The conflict will only surface when the company goes public and has to face a broader set of institutional investors who care about things like “one share, one vote.”
Let’s dive deeper into the specific dimensions of this move. From a commercialization perspective, super-voting shares allow the founders to ignore short-term sales pressure and focus on R&D. That’s consistent with Anthropic’s published stance on “responsible scaling.” But it also creates a moral hazard: if the founders make a mistake, there’s no mechanism to remove them. The board, which is already stacked with their allies, can’t act. The only check is the market—and by then, the stock has already crashed.
From a competitive landscape view, this structure positions Anthropic as a governance outlier. OpenAI uses a capped-profit model with a non-profit board. xAI is private and controlled by Musk. Google’s AI arm is part of a public company with a single-class structure (though Alphabet has a dual-class for founders). By choosing super-voting shares, Anthropic is signaling that it prioritizes founder vision over shareholder democracy. That could be a magnet for long-term investors who believe in the mission—or a red flag for those who fear founder entrenchment.
The code doesn’t care about your narrative—but it does care about the details. The key unknown is the sunset provision. Will the super-voting shares convert to ordinary shares after a period of time, say 5 or 10 years? Or after the founders’ death or incapacity? The absence of a sunset clause is the single biggest governance risk. Snap’s structure, which gives no voting rights to public shareholders, is the extreme example. Anthropic would be wise to adopt a “time-based sunset” or “trigger-based sunset” tied to the founders’ departure. That would reduce the governance discount by at least 50%, according to ISS research.
Another hidden layer: the impact on future fundraising. If Anthropic needs to raise more capital after the IPO (which is likely, given the compute costs for training next-gen models), the super-voting structure could make it harder to attract new investors. They’ll demand a higher yield or preferred shares with voting rights. That dilutes the founders’ control anyway. It’s a delicate balancing act.
From an ethical and safety standpoint, the super-voting shares are a double-edged sword. On one hand, they protect the AI safety mission from short-term profit motives. On the other hand, they concentrate power in the hands of a few individuals who may have their own biases. The AI safety community has long argued for “distributed governance” of AI—multiple stakeholders, checks and balances. Super-voting shares are the opposite: they centralize authority. If the founders decide to rush a dangerous model to market, there’s no internal mechanism to stop them. The board is powerless. The only recourse is external regulation, which is slow and reactive.
The market’s current capabilities are not equipped to price this risk. Standard valuation models ignore the possibility of catastrophic failure. But that’s exactly what makes this governance decision so critical. It’s not just about money; it’s about the entire trajectory of AI development.
Let’s turn to the regulatory angle. The SEC has been cracking down on dual-class structures, especially those with no sunset. In 2022, the SEC proposed rules requiring companies to disclose the material risks of multiple voting classes. Anthropic’s S-1 filing will be scrutinized. If the SEC forces a sunset clause, the entire plan could be derailed. The founders might have to accept a compromise: a 10-year sunset, or a provision that the super-voting shares convert if the company’s safety record is poor. That would be a positive outcome for investors.
The takeaway is clear: Anthropic’s super-voting shares are a strategic bet that the market will reward mission-driven control over short-term governance. It’s a bet that has worked for Google and Meta, but those companies had a clear path to profitability. Anthropic is still burning cash. The risk of a governance discount is real, but it might be offset by the premium that investors assign to AI safety leadership. The final answer depends on the specifics of the sunset clause and the reaction of institutional investors.
The real test will be whether Anthropic can design a sunset clause that satisfies both mission and market. Without it, the governance discount will be a permanent drag on its valuation. Watch the S-1 filing for details. The first filing will tell us everything: the voting ratio, the sunset provision, the lock-up period. Until then, the smart money is on caution. Governance is not a feature; it’s the architecture of the machine. And in AI, the architecture matters more than the code.